Content-based multi-criteria similarity assessment of CAD assembly models

The use of Digital Mock-Up (DMU) has become mainstream to support the engineering activities all along the Product Development Process. Over the years, companies generate large databases containing digital models and documents related to their products. Considering complex products, the DMU can be composed of several hundred thousand parts assembled together in assembly trees containing tens of sub-assemblies, and representing several terabytes of data. The ability to retrieve existing models is crucial for the competitiveness of companies, as it can help to leverage existing solutions, results and knowledge associated with previous products. To speed up the access to this large amount of reusable information, CAD models search approaches have been proposed, including the so-called content-based search techniques which do not rely on metadata and data organization but exploit the implicit knowledge embedded in the models. As part of a system for the retrieval of CAD assembly models, this paper introduces a set of four measures to evaluate assembly similarities according to multiple criteria. These measures are combined to assess three different levels of similarity (local, partial and global). The local measure only considers the contribution of the parts that are similar in the compared assemblies, while partial and global measures take also into account the number of similar parts compared to the total number of parts in the query and in the target model. Moreover, an ad-hoc visualization interface has been designed to clearly highlight the different similarities and to allow a fast identification of the target models. The validation of the proposed method is discussed, the dataset used to this aim is provided with the specification of the adopted ground truth and some examples of the obtained results are shown.